How Do People Evaluate Non-Ideal Budgets? Evidence from Pairwise Comparisons
Budget aggregation allows citizens to express their ideal budget allocations, which a predetermined rule combines into a collective allocation. Existing theoretical work on budget-aggregation rules often relies on assumptions about how voters evaluate allocations that differ from their ideals, yet these assumptions have received little empirical validation. We present a framework for testing hypotheses about voters' utility functions through pairwise comparisons. Our modular, open-source polling system elicits participants' ideal allocations and presents carefully generated pairs of non-ideal alternatives. Different pair-generation algorithms enable tests of specific utility models and general preference properties. We conducted over 60 polls, yielding over 3,000 valid participant-poll observations. Our results show that most choices are inconsistent with common utility models, including $\ell_1$, $\ell_2$, and Leontief. Participants also frequently violate sign and issue symmetry, evaluating increases and decreases differently and assigning different importance to deviations across issues. In contrast, more general properties, including star-shapedness, single-peakedness, and peak-linearity, receive greater empirical support. These findings suggest that standard symmetric distance-based models incompletely capture budget-allocation preferences and motivate further evaluation of more flexible utility models for practical budget aggregation.